CN105825677B - A kind of urban traffic blocking prediction technique based on improvement BML models - Google Patents

Abstract

The invention discloses a method for predicting urban traffic congestion based on an improved BML model. Firstly, an M-BML model is constructed, and the M-BML model is initialized; model; and map the traffic flow density of a certain period of time in the road network to the M-BML model in proportion; finally, the M-BML model evolves according to the No. 184 rule of the cellular automata. When the model finally enters the congestion state, Get the coordinates of the grid points where congestion occurs on the model, and finally map the coordinates obtained on the model to the real traffic road network through corresponding strategies, and get the intersections on the real road network that may be seriously congested in the next time period .

Description

A Method of Urban Traffic Congestion Prediction Based on Improved BML Model

technical field

The invention belongs to the technical fields of computer science and intelligent transportation, and relates to an urban traffic congestion prediction method based on an improved BML model, which is used for predicting traffic congestion on an actual urban traffic road network.

Background technique

With the acceleration of industrialization and urbanization in our country, the development of cities is becoming more and more saturated. The economic development of cities and people's travel life have undoubtedly brought great pressure to urban road traffic, so there is an urgent need for a road that can guide and traffic flow theory for dispatching traffic. At present, traffic flow modeling based on cellular automata has been generally recognized by the academic community, and the research on the traffic flow model (Nagel–Schreckenberg traffic model, N-S model for short) used to describe expressways has reached maturity and has been applied into the actual traffic simulation system. However, the traffic flow model (Biham–Middleton–Levine traffic model, referred to as BML model) used to describe the urban traffic network is still at the stage of theoretical research. Most scholars’ research is to analyze the phase transition principle of the BML model. And made relevant theoretical proofs. No scholars have applied the BML model to the actual urban traffic network, and carried out related intelligent scheduling and guidance.

Contents of the invention

In order to solve the above technical problems, the present invention improves on the basic BML model, establishes an M-BML model that can truly describe the actual urban traffic characteristics, and performs real-time prediction of urban traffic congestion points by running the model.

The technical scheme adopted in the present invention is: a kind of urban traffic jam prediction method based on improved BML model, it is characterized in that, comprises the following steps:

Step 1: Construct the M-BML model and initialize the M-BML model;

Step 2: Map the criss-cross road network of real urban traffic to the M-BML model according to a certain strategy; and map the traffic flow density of a certain period of time in the road network to the M-BML model in proportion;

Step 3: The M-BML model evolves according to the No. 184 rule of the cellular automata. When the model finally enters the blockage state, the coordinate value of the grid point where the blockage occurs on the model is obtained, and finally the coordinate value obtained on the model is obtained through the corresponding strategy The value is mapped to the real traffic road network, and the intersections on the real road network that may be seriously congested in the next time period are obtained.

As a preference, the construction of the M-BML model described in step 1 is to add the concept of lines on the basis of the BML model, and map the lines between the two destinations in the urban road network to the east and north respectively once, in the model The cell is the intersection of two eastbound and northbound lines, such as intersections, tunnels, overpasses, etc.; the vehicles on each line are no longer randomly disordered, but obey the density distribution of the line. The specific calculation The formula looks like this:

Among them, M represents the number of all lines in the model, N _i (( ^1≤i≤M )) represents the number of road sections contained in the ith line, and (1≤j≤N _i ) represent the traffic flow density and length of the j ^th section on the i ^th line, respectively. According to the line vehicle density obtained by the above formula, the eastbound vehicles and northbound vehicles in the cell are randomly initialized respectively; a scale is added to the BML model to obtain the M-BML model.

The operating rules of the M-BML model are:

(1) The model adopts periodic boundary conditions, so the number of vehicles on each line is conserved;

(2) The rule of the traffic lights at the intersection is that the time steps are divided into odd time steps and even time steps. Vehicles heading east at odd time steps can use it, and vehicles traveling north at even time steps can use it; at odd time steps When , a vehicle traveling east can only drive to the right when the right cell is empty; at an even time step, a vehicle traveling north can also drive upward only when the upper cell is empty;

(3) The vehicle speed can only take values between (0, 1).

As a preference, the initialization of the M-BML model described in step 1 is to obtain the vehicle density value of each road in the entire city at the set time t, and then the vehicle density value of the corresponding line calculated according to formula 1 is to the M-BML model to initialize. Different initial traffic flow densities will lead to different final states of the system, such as free flow state, intermediate state or blocked flow state. Through the experimental simulation, it is found that the traffic flow density is between 0.3 and 0.5, which is the critical interval for the system to change from the free flow phase to the blocked phase.

As a preference, the real urban traffic criss-cross road network described in step 2 is mapped to the M-BML model according to the following strategy:

(1) Select an optional path set from a starting point O to a destination D. Under the premise of not considering the situation of turning around and assuming that each road segment is selected at most once, it can be obtained according to the following steps. First, establish a search tree with the starting point O as the root, each intersection as the child node, and a certain level of expansion; secondly, with the destination D as the end point, find all the child nodes that traverse from the tree root to the end point in the search tree The formed path is the set of optional paths.

(2) Evaluate these feasible paths, and fill in the paths that meet the selection criteria into the corresponding table items about the paths between the corresponding two intersections of the urban road network. The route selection criteria include the vehicle's preference for a certain route and the traffic status of the corresponding route. The vehicle's preference for a certain path does not only depend on the distance and travel time of the path, but also considers other factors, such as some objective attributes of the road sections included in the path, including the number of lanes, whether there are crosswalks, and whether lighting equipment is sufficient etc., as well as the different subjective preferences of drivers for roads; the traffic status of a route refers to the occurrence of uncertain traffic events, etc.

(3) Map each path obtained according to the above steps to the grid of the M-BML model in the east direction and north direction respectively.

As a preference, the traffic flow density of a certain period of time in the road network described in step 2 is also mapped to the M-BML model in proportion. The length ratio of each road segment on the route is used to map the vehicle density to the M-BML model.

As preferably, described in step 3, map the coordinate value obtained on the model to the real traffic road network by corresponding strategy, be according to the type of the line intersection (such as intersection, tunnel, overpass, etc.) that grid point comprises corners, etc.), the M-BML model is mapped to the urban traffic road network, and its rules are divided into the following four points:

(1) One-to-one mapping;

If the grid point for predicting congestion only contains one intersection, then the intersection is the one where traffic congestion occurs in the real urban traffic network;

(2) Mapping of conflict points;

If the grid points that predict congestion only contain intersections and more than one, the grid points are called conflict points; through joint mapping, the grid points that are predicted to be congestion points in the same row or column are respectively subjected to intersection operations, The obtained intersection is the congestion point in the real urban road network;

(3) Mapping of fuzzy points;

If the grid points that predict congestion include overpasses, tunnels, or corners, the grid points are called fuzzy points; such cases will be ignored when the M-BML model is mapped to the real urban traffic network;

(4) Mapping of empty points;

If there are no intersections, tunnels, overpasses or corners in the predicted grid point, the grid point is called an empty point; such cases will be ignored by the M-BML model when it is mapped to the real urban traffic network.

As preferably, described in step 3 obtains the intersection that serious congestion may occur in the next period of time on the real road network, and its specific implementation process includes the following sub-steps:

Step 3.1: Load the traffic flow density of each route;

Step 3.2: Run K time steps according to the basic rules of the BML model to capture the marker value of the intersection that initially caused the congestion;

Step 3.3: Analyze which actual traffic intersections are congested according to the acquired marker values combined with the mapping rules.

As a preference, in step 3.3, according to the obtained marker value combined with the mapping rules to analyze which actual traffic intersections are jammed, each point on the grid is marked with a jam value initialized to 0, if the point The greater the congestion value of , the greater the impact of the point on the congestion of the entire traffic network, and the point exceeding the congestion threshold is marked as a congestion point.

As a preference, the update rule of the blockage value on the grid point is:

(1) When the vehicle passes through the grid point, if the traffic is smooth and there is no stagnation, then the congestion value of the point remains unchanged;

(2) When a vehicle stops at a grid point due to a vehicle in front of it, the grid point is occupied by the vehicle and other vehicles that want to pass through the point cannot pass. This point has an impact on the blockage of the entire model. Then, by punishing this point, the congestion value of this point can be increased; if the vehicle is still stuck on this grid point in the next time step, then further increase the congestion value of this grid point;

(3) When the vehicle changes from a stagnant state to a driving state, that is, the corresponding grid point changes from a blocked state to a free state, then the point can be rewarded so that its congestion value can be reduced by multiples.

Compared with the prior art, the beneficial effect of the present invention is: the present invention combines the real urban road network structure, improves the traditional BML model, and applies the simple and efficient features of the BML model to the real-time prediction of traffic jams in the real urban road network, which can Accurately predict congested intersections in the traffic network in real time.

Description of drawings

Figure 1: Mapping diagram of the present invention from eastbound traffic flow passing through an intersection to a BML model.

Fig. 2: Urban traffic road network map of the present invention.

Fig. 3: The spatial structure diagram of the M-BML model of the present invention.

Fig. 4: The mapping diagram of the congestion point on the M-BML model of the present invention to the intersection where the congestion occurs on the real road network.

Figure 5: Solution diagram of the one-to-one mapping of the present invention.

Fig. 6: Solution diagram of conflict point mapping in the present invention.

Fig. 7: Solution diagram of fuzzy point mapping of the present invention.

Figure 8: Solution diagram of empty point mapping of the present invention.

Fig. 9: Flowchart of traffic jam prediction by M-BML model of the present invention.

Figure 10: The density value initialization diagram of each road in the urban traffic road network of the present invention.

Fig. 11: The mapping of the vehicle density on the M-BML model of each section of real traffic in the present invention.

Fig. 12: Diagrams of free flow state, intermediate state and blocked state under the M-BML model of the present invention.

Detailed ways

In order to facilitate those of ordinary skill in the art to understand and implement the present invention, the present invention will be described in further detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the implementation examples described here are only used to illustrate and explain the present invention, and are not intended to limit this invention.

At present, the analysis of the global congestion state of the BML model at home and abroad is still in theoretical research, mainly studying the phase transition from the free flow state to the congestion state and theoretical proofs. No scholars have applied the BML model to actual traffic guidance, which is similar to the BML model It has a lot to do with its own limitations. First of all, the vehicles in the BML model are all driving towards a square, and do not change direction during the driving process, which is obviously inconsistent with the actual traffic flow. Then there is a congestion state in the BML model, which is caused by the mutual interference of vehicles in two directions. This only shows that the BML model has the characteristics of describing the congestion state that will occur in real traffic, but it cannot point out that it is specifically in the city. Which traffic area is blocked, it cannot solve the actual traffic problem. Finally, when the BML model is initialized, the number of eastbound and northbound vehicles is equal and randomly distributed, which does not match the actual urban traffic conditions.

In view of the shortcomings of the BML model in simulating real traffic network congestion in the above analysis, the present invention creates an M-BML model by improving the basic BML model, and predicts the real traffic network by simulating the real urban traffic network. When and where the blocked intersection occurred. Including the two-way mapping rules between the real urban traffic network and the M-BML model, as well as the data input and evolution rules of the M-BML model.

Please see Fig. 9, a kind of urban traffic jam prediction method based on improved BML model provided by the present invention, comprises the following steps:

Step 1: Construct the M-BML model and initialize the M-BML model;

The construction of the M-BML model is based on the BML model, adding the concept of lines, and mapping the lines between the two destinations in the urban road network to the east and north respectively. The cells in the model are east and north The intersection of two lines, such as intersections, tunnels, overpasses, etc.; the vehicles on each line are no longer randomly disordered, but obey the density distribution of the line. The specific calculation formula is as follows:

Among them, M represents the number of all lines in the model, N _i (( ^1≤i≤M )) represents the number of road sections contained in the ith line, and (1≤j≤N _i ) represent the traffic flow density and length of the j ^th section on the i ^th line, respectively. According to the line vehicle density obtained by the above formula, the eastbound vehicles and northbound vehicles in the cell are randomly initialized respectively; a scale is added to the BML model to obtain the M-BML model.

The operating rules of the M-BML model are:

(1) The model adopts periodic boundary conditions, so the number of vehicles on each line is conserved;

(2) The rule of the traffic lights at the intersection is that the time steps are divided into odd time steps and even time steps. Vehicles heading east at odd time steps can use it, and vehicles traveling north at even time steps can use it; at odd time steps When , a vehicle traveling east can only drive to the right when the right cell is empty; at an even time step, a vehicle traveling north can also drive upward only when the upper cell is empty;

(3) The vehicle speed takes a value between (0, 1).

Initializing the M-BML model is to obtain the vehicle density value of each road in the whole city at the set time t, and then initialize the M-BML model according to the vehicle density value of the corresponding line calculated by formula (1). Different initial traffic flow densities will lead to different final states of the system, such as free flow state, intermediate state or blocked flow state. Through the experimental simulation, it is found that the traffic flow density is between 0.3 and 0.5, which is the critical interval for the system to change from the free flow phase to the blocked phase.

As shown in Figure 1(a), the present invention selects two driving routes from the complex urban traffic road network, which are respectively the route (route 1) from the starting point S1 to the end point E1 through two intersections and the route from the starting point S2 departs from the route (route 2) that passes through two crossroads and reaches the end point E2. It can be seen from the figure that the two driving routes are not straight lines. Route 1 passes through two intersections and turns in two directions. Route 2 also passes through two intersections and turns in two directions. There is a section of road between the two routes. overlap together.

The object of the present invention is to predict and analyze at which intersection the urban traffic road network congestion occurs, regardless of the situation that a certain section of road is blocked. As shown in the figure, the present invention only pays attention to the occurrence of congestion at two crossroads, but does not pay attention to the traffic congestion caused by traffic accidents or road narrowing on the Ri road section. Due to the limitations of the BML model itself, the present invention cannot directly copy the simple road network composed of these two routes to the BML model.

The present invention uses two routes (S1->E1S2->E2) shown in FIG. 1(b) to simulate route 1 and route 2 in real traffic, and constructs an M-BML model. If there is congestion at the intersection of the S1->E1 route and the S2->E2 route on the M-BML model, it can be predicted that intersection 1 or intersection 2 in the left figure (left turning and straight going will cross, both turn right will not intersect) there is a jam, so that the intersection where the real traffic jam occurs can be obtained through the simulation experiment of the M-BML model.

As shown in Figure 2 is a simple urban traffic road network structure diagram. This picture basically reflects the basic building facilities and road network structure of a city. It can be seen from the picture that the city's travel locations mainly include suburban residences, schools, railway stations, etc., which are marked with English capital letters A-F respectively. The city routes criss-cross to form different intersections. It can be seen from the figure that there are four three-way intersections marked C1C2C4C9, three intersections marked C3C5C8 and two bends C6C7.

Because the present invention simulates the urban traffic road network with the M-BML model to predict at which intersections the congestion will occur, so only the congestion caused by the mutual restriction of vehicles in different directions is considered, and the congestion caused by traffic accidents is not considered. Therefore, the traffic jam point predicted by the present invention should be the intersection from C1 to C9, not on the road section or travel point.

Table 1 Urban traffic route table

If you want to simulate the operation of real road network traffic through the M-BML model, you must first map the lines on the real road network to the M-BML model. As shown in Figure 2, there are five destinations that can be used from travel point A, and there are multiple routes to each destination. For example, from suburban residence A to school B, you can use the route A-C1-C2-B Driving, you can also drive on A-C1-C3-C5-C2-B, or from the line A-C1-C3-C7-C8-C5-C2-B. Here the present invention considers the driving habit of the driver and only selects the first route. As shown in Table 1, all the routes in Figure 2 are extracted, and there are a total of 30 routes. It can be seen from the table that there are many lines with the same starting and ending points, such as Line 2 and Line 3, because these two lines should be considered according to the road network conditions at the time from the driver's driving habits, and the distance between the two lines is not much different. If the driver finds that Line 2 is crowded, he may drive from Line 3. As mentioned above, the reason why drivers habitually choose Line 1 from point A to point B is that other routes are far longer than Line 1.

Now consider how to map the 30 routes in Table 1 to the M-BML model. Since the basic BML model has no concept of routes, each grid point represents an intersection, and the vehicles on the model are randomly distributed. , it only shows that the congestion will be caused by the driving constraints of vehicles in both directions. Now the improvement of the M-BML model to the basic BML model is to add the concept of the line. The vehicles on each line are no longer randomly disordered, but obey the density distribution of the line. Add a scale to the model to improve the basic BML model, that is, build M-BML, as shown in Figure 3. Since the urban road network shown in Table 1 has 30 routes, the size of the model is 30×30 when mapped to the model, that is, there are 30 routes in the middle of the model and 30 routes in the north direction. The eastbound route 14 and northbound route 9 circled in the figure are the two routes B-C2-C1-C3-C4-C and A-C1-C2-C5-C6-C9-F corresponding to Table 1 , the coordinate value of their intersection point is (9, 14). At this time, it can be seen from the figure that a northbound vehicle stops at this coordinate point and is in a stagnant state. The present invention can consider that the coordinate value corresponds to There was a blockage at the intersection. Combining Table 1 and Figure 2, it can be analyzed that the intersections on the real road network corresponding to the grid point on the M-BML model are intersections C3 and C5, that is, the present invention can determine one or both of C3 and C5 All blocked.

Step 2: Map the criss-cross road network of real urban traffic to the M-BML model according to a certain strategy; and map the traffic flow density of a certain period of time in the road network to the M-BML model in proportion;

Map the criss-cross road network of real urban traffic to the M-BML model according to the following strategy:

(1) Select an optional path set from an origin (O) to a destination (D). Under the premise of not considering the situation of turning around and assuming that each road segment is selected at most once, it can be obtained according to the following steps. First, establish a search tree with the starting point (O) as the root, and each intersection as a child node, with a certain level of expansion; secondly, with the destination (D) as the end point, find all the paths from the tree root to the end point in the search tree The path formed by the child nodes of is the set of optional paths.

(2) Evaluate these feasible paths, and fill in the paths that meet the selection criteria into the corresponding table items about the paths between the corresponding two intersections of the urban road network. The route selection criteria include the vehicle's preference for a certain route and the traffic status of the corresponding route. The vehicle's preference for a certain path does not only depend on the distance and travel time of the path, but also considers other factors, such as some objective attributes of the road sections included in the path, including the number of lanes, whether there are crosswalks, and whether lighting equipment is sufficient etc., as well as the different subjective preferences of drivers for roads; the traffic status of a route refers to the occurrence of uncertain traffic events, etc.

(3) Map each path obtained according to the above steps to the grid of the M-BML model in the east direction and north direction respectively.

The traffic flow density of a certain period of time in the road network is also mapped to the M-BML model in proportion. The vehicle density of the current road section is obtained through the real-time monitor at each intersection, and the length ratio of each section of the road on the original actual route is used. To map the vehicle density to the M-BML model.

Step 3: The M-BML model evolves according to the No. 184 rule of the cellular automata. When the model finally enters the blockage state, the coordinate value of the grid point where the blockage occurs on the model is obtained, and finally the coordinate value obtained on the model is obtained through the corresponding strategy The value is mapped to the real traffic road network, and the intersections on the real road network that may be seriously congested in the next time period are obtained.

The result obtained by running the M-BML model is that the congestion points on the model are mapped to the intersections of the real road network, as shown in Figure 4. According to the adjustment of the parameter values, it is assumed that the congestion range is within the area delineated in the left figure. The coordinate values of these grid points can be obtained from the figure. For example, the coordinate value of the grid point in the lower left corner is (12, 17), indicating that this point is the intersection of Line 12 and Line 17, corresponding to the real traffic network In the C5 intersection, then it can be predicted that the C5 intersection in the real road network will be blocked in the future time period.

According to the types of line intersections contained in the grid points, the rules for mapping the M-BML model to the urban traffic network are divided into the following four points:

(1) One-to-one mapping

If the grid point for predicting congestion contains only one intersection, then this intersection is the one where traffic congestion occurs in the real urban traffic network. As shown in Figure 5, it is assumed that after the M-BML model is run, the grid point that finally predicts congestion is (14, 21) (16, 19). These two grid points correspond to one-to-one mapping, so the mapping to The congested intersections obtained in the real road network are C2 and C4.

(2) Mapping of conflict points

If the grid point that predicts congestion only contains intersections and more than one, the grid point is called a conflict point. A good solution to conflict points is through joint mapping, that is, to perform intersection operations on the grid points predicted to be congestion points in the same row or column, and the obtained intersections are the congestion points in the real urban road network.

As shown in Figure 6, assuming that the system finally determines that the three grid points with coordinate values of (14, 20) (14, 21) (16, 19) are the most congested intersections, these three coordinate points can be combined From analysis, the intersections of the real network corresponding to these three points are C3C4, C4 and C2 respectively. The grid point (16, 19) corresponds to the C2 intersection is a one-to-one mapping. The other two grid points (14, 20) (14, 21) are in the same column, and the intersection is taken to obtain that both grid points correspond to the C4 intersection, so it is determined that the C4 intersection is blocked. Combining grid points (16, 19), the final blocked intersections are C2 and C4.

(3) Mapping of fuzzy points

If a grid point predicting congestion includes an overpass, tunnel or corner, the grid point is called an ambiguous point. Because the congestion predicted by the M-BML model is mainly caused by traffic flows from directions crossing each other, and congestion at overpasses, tunnels or corners will not be considered in the model, so the M-BML model is mapped to Ignored in real city traffic network.

As shown in FIG. 7 , it is assumed that the system finally determines that the three grid points whose coordinate values are (16, 18) (17, 18) (17, 21) are the most congested intersections. The first grid point (16, 18) is a one-to-one mapping, corresponding to intersection C2. The latter two grid points (17, 18) (17, 21) both contain tunnel C8, so the tunnel is ignored, and the intersection is taken to correspond to intersection C5, so it is determined that C5 is blocked. Combining the grid points (16, 18), the final blocked intersections are C2 and C5.

(4) Mapping of empty points

If there are no intersections, tunnels, overpasses, corners, etc. in a grid point for which congestion is predicted, the grid point is called an empty point. This corresponds to the real urban road network, where the east and north directions of the same route overlap, or two disjoint routes, so when mapping the M-BML model to the urban road network, empty points will also be ignored.

As shown in FIG. 8 , it is assumed that the system finally determines that the three grid points whose coordinate values are (13, 20) (16, 18) (16, 21) are the most congested intersections. The first two grid points (13, 20) (16, 18) are all one-to-one mappings, corresponding to intersections C4 and C2, respectively. And the last grid point (16, 21) is an empty point, so the congestion grid point will be ignored during mapping. The resulting blocked intersections are C2 and C4.

The definition of the improved BML model is basically similar to that of the original BML model. It follows the same rules of traffic lights and also adopts periodic boundary conditions. The speed of the vehicle can only take a value between (0, 1). The difference lies in the model's The initialization is not random, and the density of vehicles traveling eastward is no longer the same as that of northbound vehicles, but the data is obtained from actual traffic monitoring.

The M-BML model is a mapping of real traffic network routes, and it retains most of the characteristics of the BML model. The difference is that the initialization of the M-BML model is not randomly assigned, but the density value of each road in the entire city is obtained at the set time t, and then the corresponding density value is assigned to each road in the M-BML model. On the route, the model is close to the real traffic road network.

In Figure 2, the road map of the entire urban traffic network includes intersections and intersections and intersections and intersections and intersections. The present invention also needs to measure the traffic flow density value of the starting point and the intersection when observing the traffic flow density data of the road section. Because the congestion for each intersection is affected by the traffic flow in all directions. For example, the intersection C1 may be blocked due to the excessive traffic flow values of the A-C1 road section and the C2-C1 road section, so the present invention will consider the traffic flow density on the A-C1 road section.

According to the urban traffic road network diagram in Figure 2, an initial traffic flow density value is assigned to each road segment, as shown in Figure 10. There are 5 intersections in the line from the starting point D to the end point F. Due to the influence of the intersections, the traffic density of each road section will vary greatly, so the traffic flow of each road section must be mapped to the M-BML model in sections line up, as shown in Figure 11. The route consists of 6 road sections. The vehicle density on each road section is determined according to the actual road conditions. If the vehicles are randomly distributed at the initial moment in the BML model, it cannot reflect the real road network traffic conditions. You can pass through each intersection The real-time monitor to obtain the vehicle density of the current road section. It is very easy to obtain the distribution of the density value of each route after the traffic flow density value of each road segment has been monitored, as shown in Table 2. When the densities on the six road sections in Figure 11 are mapped to the M-BML model, they should be mapped according to the length ratio of each road section on the original actual route.

Table 2 Distribution of traffic flow density values for each route in the urban traffic network

After the traffic flow of each road section on each route has been determined, the present invention randomly initializes the vehicle distribution on the M-BML model according to the probability, because it is impossible to directly copy the driving state of each vehicle in the actual traffic to the model , but it can only be said that at a certain moment the traffic flow density on the corresponding road section in the model is equal to the traffic flow density on the road section in the actual traffic.

After running through computer numerical simulation, the final state of the M-BML model system may be a free flow state, an intermediate state or a blocked flow state, as shown in Figure 12. Consider marking each point on the grid with a congestion value initialized to 0. If the congestion value of the point is larger, it means that the point has a greater impact on the congestion of the entire traffic network.

Update rules for blockage values on grid points:

(1) When the vehicle passes through the grid point, if the traffic is smooth and there is no stagnation, then the congestion value of the point remains unchanged.

(2) When a vehicle stops at a grid point due to a vehicle in front of it, the grid point is occupied by the vehicle and other vehicles that want to pass through the point cannot pass. This point has an impact on the blockage of the entire model. Then making a penalty for this point can increase the blockage value of this point. If the vehicle is still stuck on the grid point in the next time step, further increase the congestion value of the grid point.

(3) When the vehicle changes from a stagnant state to a driving state, that is, the corresponding grid point changes from a blocked state to a free state, then the point can be rewarded so that its congestion value can be reduced by multiples.

After the system runs, all grid points have different blockage values. At this time, the grid points with relatively large blockage values are selected. It can be considered that these points are the most severely blocked points in the entire model, and the grid point of this point is obtained. coordinate value.

It should be understood that the parts not described in detail in this specification belong to the prior art.

It should be understood that the above-mentioned descriptions for the preferred embodiments are relatively detailed, and should not therefore be considered as limiting the scope of the patent protection of the present invention. Within the scope of protection, replacements or modifications can also be made, all of which fall within the protection scope of the present invention, and the scope of protection of the present invention should be based on the appended claims.

Claims (8)

1. a method for predicting urban traffic congestion based on improved BML model, is characterized in that, comprises the following steps: Step 1: Construct the M-BML model and initialize the M-BML model; Step 2: Map the criss-cross road network of real urban traffic to the M-BML model according to a certain strategy; and map the traffic flow density of a certain period of time in the road network to the M-BML model in proportion; Step 3: The M-BML model evolves according to the No. 184 rule of the cellular automata. When the model finally enters the blockage state, the coordinate value of the grid point where the blockage occurs on the model is obtained, and finally the coordinate value obtained on the model is obtained through the corresponding strategy The value is mapped to the real traffic road network, and the intersections on the real road network that may be seriously congested in the next time period are obtained; The coordinate values obtained on the model are mapped to the real traffic road network through the corresponding strategy. According to the type of line intersections contained in the grid points, the M-BML model is mapped to the urban traffic road network. The rules are divided into The following four points: (1) One-to-one mapping; If the grid point for predicting congestion only contains one intersection, then the intersection is the one where traffic congestion occurs in the real urban traffic network; (2) Mapping of conflict points; If the grid points that predict congestion only contain intersections and more than one, the grid points are called conflict points; through joint mapping, the grid points that are predicted to be congestion points in the same row or column are respectively subjected to intersection operations, The obtained intersection is the congestion point in the real urban road network; (3) Mapping of fuzzy points; If the grid points that predict congestion include overpasses, tunnels, or corners, the grid points are called fuzzy points; such cases will be ignored when the M-BML model is mapped to the real urban traffic network; (4) Mapping of empty points; If there are no intersections, tunnels, overpasses or corners in the predicted grid point, the grid point is called an empty point; such cases will be ignored by the M-BML model when it is mapped to the real urban traffic network. 2. the urban traffic jam prediction method based on improved BML model according to claim 1, it is characterized in that, described in step 1 builds M-BML model, is on the BML model basis, adds the concept of line, city road The lines between the two destinations in the network are respectively mapped once to the east and to the north, and the cells in the model are the intersections of the two lines to the east and to the north; the vehicles on each line are no longer randomly disordered, Instead, it obeys the density distribution of the line, and the specific calculation formula is as follows: Among them, M represents the number of all lines in the model; N _i represents the number of sections contained in the ith line, ^1≤i≤M ; and represent the traffic flow density and length of the j ^th road section on the i ^th line respectively, 1≤j≤N _i ; According to the line vehicle density obtained by the above formula, randomly initialize the eastbound vehicles and northbound vehicles in the cell respectively; add a scale to the BML model to obtain the M-BML model; The operating rules of the M-BML model are: (1) The model adopts periodic boundary conditions, so the number of vehicles on each line is conserved; (2) The rule of the traffic lights at the intersection is that the time steps are divided into odd time steps and even time steps. Vehicles heading east at odd time steps can use it, and vehicles traveling north at even time steps can use it; at odd time steps When , a vehicle traveling east can only drive to the right when the right cell is empty; at an even time step, a vehicle traveling north can also drive upward only when the upper cell is empty; (3) The vehicle speed takes a value between (0, 1). 3. the urban traffic jam prediction method based on improved BML model according to claim 2, is characterized in that, described in the step 1, initializes M-BML model, is to obtain every road of the whole city at the moment t of setting The vehicle density value, and then the M-BML model is initialized according to the vehicle density value of the corresponding line calculated by formula 1. 4. the urban traffic congestion prediction method based on improved BML model according to claim 1, it is characterized in that, described in step 2, the road network that real urban traffic crisscrosses is mapped on the M-BML model according to the following strategies: (1) Select a set of optional paths from a starting point O to a destination D; without considering the situation of turning around and assuming that each road section is selected at most once, it can be obtained according to the following steps; first, establish a tree with the starting point O as The root and each intersection are child nodes, and a search tree with a certain level of expansion; secondly, with the destination D as the end point, it is optional to find out in the search tree all the child nodes that traverse from the tree root to the end point. collection of paths; (2) Evaluate these feasible paths, and fill in the paths that meet the selection criteria into the corresponding table items about the path between two intersections in the urban road network; the path selection criteria include the vehicle's preference for a certain path and the corresponding path traffic status; (3) Map each path obtained according to the above steps to the grid of the M-BML model in the east direction and north direction respectively. 5. the urban traffic congestion prediction method based on improved BML model according to claim 1, is characterized in that, the traffic flow density of a certain period of time in the road network described in step 2 is also mapped in the M-BML model according to proportion, The vehicle density of the current road section is obtained through the real-time monitor of each intersection, and the vehicle density is mapped to the M-BML model according to the length ratio of each road section on the original actual route. 6. according to the described urban traffic congestion prediction method based on improved BML model according to any one of claim 1-5, it is characterized in that, described in step 3 obtains the intersection that serious congestion may take place in the next time period on the real road network intersection, its specific implementation process includes the following sub-steps: Step 3.1: Load the traffic flow density of each route; Step 3.2: Run K time steps according to the basic rules of the BML model to capture the marker value of the intersection that initially caused the congestion; Step 3.3: Analyze which actual traffic intersections are congested according to the acquired marker values combined with the mapping rules. 7. the urban traffic jam prediction method based on improved BML model according to claim 6, it is characterized in that: described in the step 3.3 according to the mark value that has obtained in conjunction with mapping rule to analyze specifically which actual traffic crossings have jammed, It is to mark each point on the grid with a congestion value initialized to 0. If the congestion value of the point is greater, it means that the point has a greater impact on the congestion of the entire traffic network. Points exceeding the congestion threshold are marked as congestion points . 8. the urban traffic congestion prediction method based on improved BML model according to claim 7, is characterized in that, the update rule of the blocking value on the grid point is: (1) When the vehicle passes through the grid point, if the traffic is smooth and there is no stagnation, then the congestion value of the point remains unchanged; (2) When a vehicle stops at a grid point due to a vehicle in front of it, the grid point is occupied by the vehicle and other vehicles that want to pass through the point cannot pass. This point has an impact on the blockage of the entire model. Then make a penalty for this point to increase the congestion value of this point; if the vehicle is still stuck on this grid point in the next time step, further increase the congestion value of this grid point; (3) When the vehicle changes from a stagnant state to a driving state, that is, the corresponding grid point changes from a blocked state to a free state, then the point is rewarded so that its congestion value is reduced by multiples.